Feature-Guided Search Fireworks Algorithm
Liping Yang, Dan Wang · 2025
A Feature-Guided Search Fireworks Algorithm (FSFWA) is proposed, which primarily introduces a guided search strategy based on Linear Discriminant Analysis (LDA). This strategy leverages LDA to extract the features of elite sparks, generating guiding operators to optimize the population's search direction. This approach addresses the limitations of the standard Fireworks Algorithm (FWA), which relies solely on random search, thereby enhancing the algorithm's convergence speed. Additionally, to balance local and global searches, a dynamic explosion search strategy based on population entropy is proposed, followed by a competitive selection strategy to choose the offspring. Comparative experiments on multiple international standard test functions demonstrate that the Feature-Guided Search Fireworks Algorithm exhibits superior optimization performance in terms of both search speed and accuracy compared to the traditional Fireworks Algorithm.